Caine Prize winner denies AI authorship as organizers review detector-based allegation
The award remains intact. Its chair rejects AI-detection software as a final judge, while the journalist behind the allegation cites accuracy research and acknowledges limited independent testing.
Listen to this story
The audio brief
Story brief
3 key pointsThe dispute now turns on whether an AI detector score can support a literary-integrity finding: Pangram’s classification was reported by critic Vauhini Vara, not verified evidence of the writer’s process. Caine Prize organizers say they have opened a human-centered review, while chair Ellah Wakatama rejects treating detection software as the final arbiter. The $10,000 award remains in place, and the case highlights the gap between detector-company claims and limited independent research on accuracy.
- 01
Kundizeza says the story is entirely his own work, passed through standard editorial processes, and draws on his family’s experiences during Zimbabwe’s economic turmoil.
- 02
Vara reported that Pangram classified almost all of the story as AI-generated, with an epigraph attributed to poet Vangile Gantsho as the exception.
- 03
Vara cited Pangram research and a study of academic papers, but acknowledged there is relatively little third-party research on detector accuracy.
Zimbabwean writer Simbarashe Kundizeza has denied using generative AI to write “The Blue Fighter,” as Caine Prize organizers review an allegation against their newly named winner. He defended his authorship in an October 14 interview with the Associated Press, following a public denial the previous night. The $10,000 award has not been withdrawn.
The story was announced as the Caine Prize winner on October 6, 2026. Five days later, author and journalist Vauhini Vara published an Atlantic article saying that Pangram, an AI-detection tool, had classified almost all of it as AI-generated. Kundizeza says the story is entirely his own work.
A detector result and a judgment about style
According to the Guardian’s account of Vara’s article, she ran the story through Pangram, which itself uses AI. She reported that the exception to its almost entirely AI-generated assessment was an epigraph attributed to an unpublished poem by South African writer Vangile Gantsho. That is the detector result Vara reported, not an established finding about Kundizeza’s writing process.
Vara also described features she associated with AI-produced prose: sentences built around a “not x, but y” contrast, abundant metaphors and platitudinous aphorisms. Her criticism went beyond how the text might have been produced. She called the story maudlin and full of tired tropes, while acknowledging that literary taste is subjective.
To support her reliance on the software, Vara pointed to Pangram’s own research, which found false positives were rare, including for non-native English speakers. She also cited a peer-reviewed article that found strong accuracy on mock academic papers written fully by AI, partly with AI or without it.
Those examples concern different kinds of evidence: the detector company’s research and an academic-paper test. In an interview with the BBC’s Front Row, Vara acknowledged that there was “relatively little third-party research” on AI detectors’ accuracy. Her cited evidence therefore came with an explicit limitation on independent scrutiny.
Kundizeza’s answer: personal history and a changing craft
Kundizeza’s defense starts with an unqualified denial of generative AI use. In his statement on X, he said he wrote the story himself and intended to continue working with the Caine Prize team. He also said the work had gone through standard editorial processes, according to AP.
He described “The Blue Fighter” as a deeply personal story about economic collapse and family relationships. Characters and the setting drew on events in his own life, he said. His account ties the fiction to his family’s experience during Zimbabwe’s economic turmoil, rather than to software-generated material.
The two writers also disagree about what a change in style means. Vara said a Kundizeza story from 2021 differed tonally and stylistically from the winning entry. Kundizeza rejected the premise that recurring interests require a fixed template, arguing that writers can change genres and improve substantially over five or more years.
In his AP interview, Kundizeza raised a further concern: technology could be deployed in ways that create bias against people writing about their own experiences and cultures. He called for responsible use of those tools and urged African writers and readers to challenge such accusations. That is his concern, not a demonstrated finding of bias in this case.
I am not happy for the work of writers, and in this case especially black writers, to form the testing ground for something that is at best still rough and ready.
Ellah Wakatama, Caine Prize chair, speaking to BBC Front Row, as quoted by the Guardian
A review without a verdict
The prize organizers say they have a responsibility to examine allegations when they arise. They have begun what they describe as a human-centered process, with support and care for writers during the review. Their response commits to an investigation without treating the accusation itself as grounds to retract the prize.
Wakatama drew a sharper line on Front Row. She said she was not against AI, but opposed using it as a final arbiter of this dispute. She accused the Atlantic of already declaring the winner a cheater without properly investigating. Her objection concerns the authority given to the detector, not simply whether prize organizers should ask questions.
Sources
- apnews.comZimbabwe’s Caine Prize winner rejects claim he used AI to write his story
- theguardian.comZimbabwean author Simbarashe Kundizeza denies using AI to write prize-winning story
Reader comments
Newest comments first. Replies stay oldest first.